Mass Distribution of Azithromycin to Reduce Child Mortality

Summary

Published: December 2019

What is the problem?

In 2018, 5.3 million children died before the age of five. The highest rates of under-five mortality are found in sub-Saharan Africa, which account for roughly half of all under-five mortality.1 Respiratory infection, diarrheal diseases and malaria are among the primary causes of deaths for children under five in sub-Saharan Africa.2

What is the program?

Azithromycin is an antibiotic that is used to treat a variety of infections.3 It is recommended by the World Health Organization (WHO) for the treatment of trachoma, a disease of the eye caused by bacterial infection.4

Azithromycin may also have an impact on child mortality through its effect on respiratory infections, diarrhea and malaria.5 This effect is hypothesized to occur via azithromycin's ability to stop the growth of bacteria, such as those that cause infections leading to diarrhea and respiratory infections, and its activity against the parasites that cause malaria.6 However, the WHO has not approved mass use of azithromycin for this purpose.

In light of its potential impact on primary causes of child mortality, recent studies have sought to understand the impact of mass drug administration (MDA) of azithromycin for children. While it is not clear what implementation will look like in non-study contexts, existing studies use the following implementation approaches:

• Azithromycin is distributed to children, ranging from 1 month to 12 years old across studies, regardless of pre-existing conditions, unless there are specific contraindications (e.g., allergies to azithromycin).7
• Azithromycin is provided twice, or sometimes once, annually.8
• Azithromycin is generally provided to children meeting age restrictions in all households in a community, but at least one study provided azithromycin to only some households in a community.9
• Azithromycin may be added on to an existing MDA or administered on its own.10

What is the evidence of effectiveness?

The most credible evidence for the impact of azithromycin comes from a large-scale RCT (the MORDOR Trial). This trial finds azithromycin leads to substantial decreases in all-cause child mortality. Smaller-scale trials also find evidence for azithromycin's impact on mortality. Heavy reliance on a single large-scale trial and lack of clarity on the exact mechanisms for impact give us some uncertainty about the extent to which these results will generalize to new settings. We are also highly uncertain about the potential unintended negative consequences of antibiotic resistance caused by mass distribution of azithromycin.

Evidence from the MORDOR Trial

In the MORDOR Trial, 1,533 communities across three countries (Malawi, Niger, and Tanzania) were randomized into either an azithromycin or placebo group. In azithromycin communities, all individuals identified in twice-yearly censuses aged 1 to 59 months old received azithromycin twice per year over two years. The trial included approximately 190,000 children at baseline and spanned 2014-2017.11 The main results are reported in Keenan et al. 2018, though there have been several follow-up studies, described below.

The trial found all-cause mortality was 13.5% (95% CI = 6.7% to 19.8%) lower in azithromycin communities than placebo communities when pooling results across countries.12 This intent-to-treat effect was the pre-specified primary outcome.13

Our assessment finds the quality of this study to be high, based on the large sample size, blinding of treatment assignment,14 pre-registration of the trial and key outcome,15 and small reported differences in attrition across azithromycin and placebo groups.16 The study does not, however, report on any baseline differences in mortality between azithromycin and placebo communities, which raises some concerns that results could be driven by chance differences between communities at baseline.

Additional evidence on all-cause child mortality effects from azithromycin

We identified three additional RCTs that measure the impact of mass azithromycin distribution on all-cause child mortality. We identified these trials based on a quick review of articles cited in Keenan et al. 2018 and recent articles citing Keenan et al. 2018. It is possible we have missed other trials meeting these criteria, and we have reviewed these trials in considerably less detail than we reviewed Keenan et al. 2018.

Two out of three trials we identified are smaller trials that provide support for an impact of azithromycin on all-cause child mortality. The third studies the interaction of azithromycin with seasonal malaria chemoprevention (SMC), so it is not clear how applicable it is to understanding the impact of azithromycin alone.

Overall, these trials increase our confidence in the effects found in the MORDOR Trial, but we assign them a smaller weight, due to their smaller sample sizes and differences in implementation.

• Porco et al. 2009 report on a smaller-scale RCT (18,415 children aged 1-9 across 48 communities) conducted in Ethiopia from 2006-2007, known as the TANA Trial. They find that children aged 1 to 9 in communities randomly assigned to receive azithromycin had 49% (95% CI = 29%-90%) lower likelihood of all-cause mortality than children in communities that randomly had treatment delayed one year.17 Results are similar for children aged 1 to 5.18 We put a smaller weight on this trial, relative to the MORDOR Trial, because of its smaller sample size.
• O'Brien et al. 2018 report on a secondary analysis of results from an RCT (part of a larger study known as the PRET Trial) in 48 communities in Niger from 2010-2013 that compared mortality effects of annual to biannual azithromycin distribution.19 They find 19% (95% CI = 66%-100%) lower all-cause mortality among children 6 months to 5 years old in communities assigned to biannual distribution vs. annual distribution (p = 0.07). We put a smaller weight on this trial, relative to the MORDOR trial, because a) it compares different frequencies of distribution of azithromycin, rather than azithromycin vs. no treatment, and b) it has a smaller sample size.20
• Chandramohan et al. 2019 report on an RCT in Burkina Faso and Mali from 2014-2016 that randomized households to receive azithromycin and SMC (sulfadoxine-pyrimethamine plus amodiaquine) vs. SMC alone.21 They find no statistically significant differences in mortality among children aged 3 to 59 months in households across groups.22 We put a smaller weight on this trial, relative to the MORDOR Trial, for assessing the impact of azithromycin on mortality because it is unclear whether the results are driven by the effectiveness of azithromycin or by interactions between azithromycin and SMC.

For our cost-effectiveness calculations, we use the effect size estimated by Oldenburg et al. 2019, which pools results across the MORDOR Trial (Keenan et al. 2018), TANA Trial (Porco et al. 2009) and PRET Trial (O'Brien et al. 2018). Oldenburg et al. 2019 find all-cause mortality is 14.4% lower in communities that received azithromycin compared to control communities.23

We exclude Chandramohan et al. 2019 from our effect size calculation because it tested the interaction of azithromycin with SMC. However, we do incorporate this finding into our consideration of external validity below and view our cost-effectiveness analysis as applying only to settings with relatively low coverage of SMC.

Mechanisms for azithromycin's impact on mortality

We have a moderate level of uncertainty about the mechanisms through which azithromycin affects all-cause child mortality.

The MORDOR Trial (Keenan et al. 2018), TANA Trial (Porco et al. 2009) and PRET Trial (O'Brien et al. 2018) did not directly investigate the mechanisms through which azithromycin reduces all-cause mortality.24

Before the MORDOR Trial, several experts predicted that mass distribution of azithromycin would lead to a reduction in all-cause mortality and that this would be driven by azithromycin's activity against respiratory infection, diarrhea and malaria.25 In the MORDOR Trial, verbal autopsies of a random sample of 250 deaths found that the most common causes of death were malaria, diarrhea or dysentery, and pneumonia (caused by a respiratory infection).26 In the TANA Trial and PRET Trial, these causes are also mentioned as primary causes of death.27

We view this as indirect evidence for azithromycin's potential mechanisms of mortality reduction. To provide direct evidence on mechanisms, we would be most persuaded by studies showing that children randomly assigned to receive azithromycin had lower mortality or morbidity related specifically to respiratory infection, diarrhea or malaria, lower levels of bacteria known to cause respiratory infection or diarrhea, or lower clinical malaria incidence, relative to children who received a placebo. The above trials do not provide this evidence.

The only study we have found to present direct evidence for mechanisms is a follow-up study to the MORDOR Trial which compared prevalence of malaria parasites among Nigerien children in azithromycin and placebo communities. It found lower malaria prevalence among those in azithromycin communities than those in placebo communities.28 Previous studies of mass distribution of azithromycin for trachoma control have also found evidence for an impact on infections of the sort hypothesized to be driving the impacts of the MORDOR Trial.29 However, we have done only a superficial scan of this work, and we do not know how reliable it is or how representative it is of the available research on this topic. It is possible we have overlooked studies providing this direct evidence for mechanisms underlying azithromycin's impact on all-cause mortality.

Finally, as discussed above, Chandramohan et al. 2019 find no difference in mortality in households randomly assigned to receive SMC alone vs. SMC plus azithromycin. This is consistent with azithormycin's impact being driven by activity against malaria, but it is also consistent with other mechanisms.30

External validity and mediators of impact

We have moderate uncertainty about the extent to which the findings of the MORDOR trial will generalize to new settings.

In the MORDOR trial, across the three countries included in the analysis, impacts are statistically significant only for Niger, which saw an 18.1% difference in all-cause mortality between azithromycin and placebo communities (95% CI = 10%-26%). The effect was smaller and not statistically significant in Malawi (5.7 percent, 95% CI = -9%-19%) and Tanzania (3.4 percent, 95% CI = -21%-23%). However, differences across countries were not statistically significant.31

Understanding the mediators of the effect of mass distribution of azithromycin will help predict how likely the impacts found in the MORDOR Trial are to generalize to new settings. Researchers have proposed several potential mediators of the impact of mass distribution of azithromycin: baseline all-cause mortality rate, interactions between azithromycin and additional drugs administered concurrently, age of individuals receiving azithromycin, baseline mortality due to specific conditions hypothesized to be addressed by azithromycin, coverage rates within the community, number of years since the first administration of azithromycin, and annual vs. biannual administration.

The available evidence on these mediators is summarized below. Overall, we believe the strongest evidence on mediating factors comes from differences in baseline all-cause mortality rate and interactions between azithromycin and additional drugs administered concurrently. The other proposed mediators have been suggested by experts and seem plausible to us, but we are not aware of compelling empirical support for them. There may also be other mediators we have not included in this list.

• Baseline all-cause mortality rate. A follow-up analysis to the MORDOR Trial found that, among participating communities, the effect of azithromycin was higher for individuals with higher baseline mortality risk, based on both geographic location and age.32 A complementary analysis that also included data from two other trials measuring the impact of azithromycin on mortality found a relationship in the same direction (though it was not statistically significant).33 These findings suggest the impact of azithromycin is non-linear, i.e., settings with higher baseline mortality rate see a higher percentage reduction in mortality.34
• Interactions with additional drugs administered concurrently. If azithromycin works through its effect on respiratory infection, diarrhea and malaria, then it may be less effective in settings where children have received alternative treatments for these diseases. As discussed above, Chandramohan et al. 2019 show no difference in impact on mortality between SMC alone vs. SMC plus azithromycin. This suggests azithromycin will have lower impacts in areas where children also receive SMC.
• Age of individuals receiving azithromycin. In the MORDOR Trial, the effect size was highest for children age 1-5 months (25%, 95% CI = 10.6% to 36%), but the difference between age groups was not statistically significant.35
• Baseline mortality due to specific conditions hypothesized to be addressed by azithromycin. Impacts may be higher in areas that have higher mortality due to specific conditions like respiratory infection, diarrhea and malaria that have been hypothesized to be addressed by azithromycin. We could not find any studies on how mortality effects are mediated by mortality due to these specific conditions. We do believe the evidence on mechanisms discussed in the previous section constitutes some suggestive evidence that impacts will be larger for areas with higher prevalence of these conditions.
• Coverage rates within the community. Azithromycin distributions covering all individuals in a community may provide added benefits through "herd effects," relative to distributions covering only a subset of individuals (e.g., individuals of certain ages). In the MORDOR trial, there was no statistically significant link between treatment coverage and effect size.36
• Number of years since first distribution. In the MORDOR trial, effect size increased over the two years of implementation, but the differences were not statistically significant.37
• Annual vs. biannual distribution. As discussed above, the PRET Study compares annual vs. biannual distribution directly and finds a 19% higher impact on mortality in biannual vs. annual communities.38

External validity adjustments are discussed further in ourcost-effectiveness analysis of mass distribution of azithromycin for children.

Potential offsetting/negative impacts

Broad use of azithromycin may lead to macrolide-resistant bacteria, which could offset azithromycin's impact on mortality. We have only done a superficial review of antibiotic resistance from mass azithromycin distribution.

The WHO considers azithromycin to be a critically important antibiotic with a high risk of bacterial resistance.39 A recent meta-analysis, which we have not vetted, concludes that azithromycin use leads to macrolide-resistant E. coli, a cause of diarrhea in children, and S. pneumoniae, which causes pneumonia.40 Analysis of samples from children in the MORDOR trial also found that macrolide resistance was higher for children in the azithromycin communities than for those in placebo communities.41

There appears to be limited data that would permit us to estimate how this resistance would map into offsetting mortality effects. A follow-up to the MORDOR Trial tests whether the effect of azithromycin is weaker in Nigerien communities that received azithromycin previously, relative to communities randomly selected to receive azithromycin for the first time. It finds that communities receiving azithromycin for the first time had similar mortality to communities receiving azithromycin for the third year, which suggests the effect of azithromycin did not weaken in communities that had been exposed to azithromycin for two years previously.42

We consider the likelihood of bacterial resistance and the cost of resistance should it develop to be highly uncertain and difficult to model. We plan to investigate this issue in order to refine our understanding of these parameters before recommending funding, but in the interim we consider the risk of long-term harm due to antimicrobial resistance to be non-negligible.

How cost-effective is it?

Based on a cost-effectiveness model we put together in November 2019 and updated with our most recent moral weights as of January 2021, we estimate that mass distribution of azithromycin to children is in the range of cost-effectiveness of the opportunities that we expect to direct marginal donations to (about 10x cash or higher, as of 2021).43 However, our estimate relies on several assumptions about which we have a high degree of uncertainty.

Note that our cost-effectiveness analyses are simplified models that do not take into account a number of factors. There are limitations to this kind of cost-effectiveness analysis, and we believe that cost-effectiveness estimates such as these should not be taken literally due to the significant uncertainty around them. We provide these estimates (a) for comparative purposes and (b) because working on them helps us ensure that we are thinking through as many of the relevant issues as possible.

Key uncertainties for the cost-effectiveness analysis are:

• External validity adjustment. Evidence for the effect of mass distribution of azithromycin relies heavily on a single study (the MORDOR Trial). For this reason, and because the mechanisms that might drive and mediate the impact of azithromycin in new settings are still unclear, we have some uncertainty about the likely effect of mass distribution of azithromycin in new settings. Two other mediators of impact across settings, discussed above, are baseline mortality and SMC coverage. Our best guess (or "preferred") cost-effectiveness estimate is based on implementation in settings similar to those studied in the MORDOR Trial—namely, settings with similar levels of baseline mortality and SMC coverage. We would not expect to see the same level of cost-effectiveness in settings with a low baseline mortality rate or in areas with high coverage of SMC, where mass distribution of azithromycin would likely have much lower effects.
• Offsetting effects of antibiotic resistance. We are highly uncertain about the negative impacts of antibiotic resistance caused by mass distribution of azithromycin, which could potentially lower the cost-effectiveness of this intervention. We have not yet attempted to calculate how much antibiotic resistance would impact our bottom-line cost-effectiveness results.
• Program costs and leveraging/funging. We have little information about program costs. We have estimated they are in the same range as the costs of vitamin A supplementation but are uncertain about this. As a result, we are also unsure about the extent to which potential leveraging and funging effects will impact our cost-effectiveness analysis.
• Development effects. For other early-life health interventions that are implemented by our top charities, we have incorporated effects due to the interventions' impacts on development. For consistency, we have included in our CEA development effects in a similar range to those of vitamin A supplementation, SMC and bednets. However, we are uncertain about this parameter, given the lack of studies on these effects for azithromycin and uncertainty about the mechanisms through which azithromycin impacts mortality (i.e., whether they are similar to the mechanisms through which SMC or vitamin A supplementation influence mortality).

The cost-effectiveness analysis explores how sensitive results are to adjustments in key parameters that would make azithromycin more or less cost-effective.

Is there room for more funding?

We are highly uncertain about the need for more funding. Child mortality remains a problem, especially in sub-Saharan Africa, suggesting a high need for interventions that aim to address child mortality. We have not explored which organizations would be willing and able to implement this intervention, though it is likely organizations that administer other drug regimens (e.g., vaccines, vitamin A supplementation, SMC) could deliver azithromycin as well. Based on a private conversation, we know there are funders who may be interested in supporting the scale-up of this type of program, but we have not dug deeper to understand the extent to which this funding would meet demand.

Key questions for further investigation

• How strong is the evidence that azithormycin's impact on all-cause child mortality is driven by its activity against respiratory infection, diarrhea and malaria?
• How likely is it that antibiotic resistance as a result of mass distribution of azithromycin would offset impacts on all-cause mortality in the short and long term?
• What are the costs of implementing mass distribution of azithromycin?
• What do ongoing studies on the impact of azithromycin find, and how do they update our estimates of the effect of this intervention?
• How do other characteristics of the intervention mediate its impact (e.g., community- vs. individual-level treatment creating "herd" effects, administering to younger vs. older children)?
• Is scale-up of mass distribution of azithromycin likely to occur in settings where impacts may be higher (e.g., areas with higher baseline all-cause mortality)?
• How large are the impacts on morbidity related to respiratory infection, diarrhea and malaria, as well as trachoma?
• How large are additional potential adverse events associated with mass distribution of azithromycin?
• What organizations would implement this program?

Sources

Document	Source
Arzika et al. 2019	Source
Chandramohan et al. 2019	Source
Coles et al. 2012	Source
Doan et al. 2019a	Source
Doan et al. 2019b	Source
Fry et al. 2002	Source
GiveWell, Cost-effectiveness analysis, 2020, Version 2	Source
GiveWell, Cost-effectiveness analysis of mass distribution of azithromycin, 2021	Source
Global burden of disease study 2017	Source
Keenan et al. 2018	Source
Keenan et al. 2018, Supplementary appendix	Source
Keenan et al. 2019	Source
O'Brien et al. 2018	Source
O'Brien et al. 2019	Source
Oldenburg et al. 2019	Source
Oron et al. 2019	Source
Porco et al. 2009	Source
Porco et al. 2019	Source
See et al. 2015	Source
UNICEF 2018	Source
WebMD, Azithromycin tablet macrolide antibiotics	Source
WHO, Trachoma, 2019	Source
WHO, Under-five mortality	Source
WHO, The Selection and Use of Essential Medicines, 2017	Source

• 1
  • "5.3 million children under age five died in 2018. The risk of a child dying before completing five years of age is still highest in the WHO African Region (76 per 1000 live births), around 8 times higher than that in the WHO European Region (9 per 1000 live births)." WHO, Under-five mortality.
  • See Table 1, UNICEF 2018, p. 8.
  • "With shifting demographics, the burden of child deaths is heaviest in sub-Saharan Africa. The burden of child deaths varies geographically, with most deaths taking place in just two regions. In 2017, half of the deaths among children under age 5 occurred in sub-Saharan Africa, and another 30 per cent occurred in Southern Asia." UNICEF 2018, p. 8.

• 2

  In sub-Saharan Africa, lower respiratory infections (including pneumonia), diarrheal diseases and malaria accounted for 14%, 22% and 16% of deaths for children ages 1-4, according to our analysis of IHME's @Global Burden of Disease Study 2017@. Calculations here.

• 3

  "Azithromycin is used to treat a wide variety of bacterial infections. It is a macrolide-type antibiotic. It works by stopping the growth of bacteria." WebMD, Azithromycin tablet macrolide antibiotics.

• 4

  WHO, Trachoma, 2019:

  • "Trachoma is a disease of the eye caused by infection with the bacterium Chlamydia trachomatis.
  • "Elimination programmes in endemic countries are being implemented using the WHO-recommended SAFE strategy. This consists of:
    • Surgery to treat the blinding stage (trachomatous trichiasis);
    • Antibiotics to clear infection, particularly mass drug administration of the antibiotic azithromycin, which is donated by the manufacturer to elimination programmes, through the International Trachoma Initiative;
    • Facial cleanliness; and
    • Environmental improvement, particularly improving access to water and sanitation."

• 5

  "Some degree of mortality reduction is plausible, since the antibiotic has activity against the three most common causes of childhood mortality in the developing world: malaria, diarrhea, and respiratory infections." See et al. 2015, p. 1.

• 6
  • "Before the trial, experts thought that a protective effect would most likely be due to reductions in respiratory infections, diarrhea, and malaria (in that order). Such a hypothesis seems reasonable, given the activity of azithromycin against bacterial pathogens of the lungs and gastrointestinal tract and against the Plasmodium falciparum apicoplast." Keenan et al. 2018, p. 1590.
  • "In Ethiopia, azithromycin is likely effective against the major pathogenic causes of lower respiratory tract infections such as Streptococcus pneumoniae and Haemophilus influenzae, and may have some effect against major causes of bacterial diarrhea such as Escherichia coli and Clostridium jejuni. Azithromycin has also been shown to have efficacy in the prevention and treatment of malaria due to both Plasmodium falciparum and Plasmodium vivax." Porco et al. 2009, p. 967.

• 7
  • "Children 1 to 59 months of age were identified in twice-yearly censuses and were offered participation in the trial." Keenan et al. 2018, p. 1583.
  • "Children who were known to be allergic to macrolides were not given azithromycin or placebo." Keenan et al. 2018, p. 1585.
  • "Objective: To compare mortality rates of participants aged 1 to 9 years in treated communities with those in untreated communities." Porco et al. 2009, p. 962.
  • "Methods: In the PRET [Partnership for the Rapid Elimination of Trachoma] cluster-randomized trial in Niger, 24 communities were randomized to annual treatment of everyone and 24 communities were randomized to biannual treatment of children under 12 for 3 years (clinicaltrials.gov, NCT00792922). Treatment was a single dose of directly observed oral azithromycin (20mg/kg up to 1gm in adults). Vital status was assessed during annual census and monitoring visits. In this pre-specified secondary analysis, we compared the mortality rate among children 6 months to less than 5 years of age by treatment arm using negative binomial regression." O'Brien et al. 2018, p. 1082.

• 8
  • "In this cluster-randomized trial, we assigned communities in Malawi, Niger, and Tanzania to four twice-yearly mass distributions of either oral azithromycin (approximately 20 mg per kilogram of body weight) or placebo." Keenan et al. 2018, p. 1583.
  • "Methods: In the PRET cluster-randomized trial in Niger, 24 communities were randomized to annual treatment of everyone and 24 communities were randomized to biannual treatment of children under 12 for 3 years (clinicaltrials.gov, NCT00792922). Treatment was a single dose of directly observed oral azithromycin (20mg/kg up to 1gm in adults). Vital status was assessed during annual census and monitoring visits. In this pre-specified secondary analysis, we compared the mortality rate among children 6 months to less than 5 years of age by treatment arm using negative binomial regression." O'Brien et al. 2018, p. 1082.

• 9
  • "A total of 1533 communities underwent randomization, 190,238 children were identified in the census at baseline, and 323,302 person-years were monitored. The mean (±SD) azithromycin and placebo coverage over the four twice-yearly distributions was 90.4±10.4%." Keenan et al. 2018, p. 1583.
  • "A limitation of the trial is that randomization was performed according to household rather than village; randomization according to household reduced the potential for bias but precluded the potential for a herd effect that might have occurred had randomization according to village been performed." Chandramohan et al. 2019, p. 2205.

• 10

  "We randomly assigned children 3 to 59 months of age, according to household, to receive either azithromycin or placebo, together with sulfadoxine–pyrimethamine plus amodiaquine, during the annual malaria-transmission season in Burkina Faso and Mali." Chandramohan et al. 2019, p. 2197.

• 11

  Keenan et al. 2018:

  • "Methods: In this cluster-randomized trial, we assigned communities in Malawi, Niger, and Tanzania to four twice-yearly mass distributions of either oral azithromycin (approximately 20 mg per kilogram of body weight) or placebo. Children 1 to 59 months of age were identified in twice-yearly censuses and were offered participation in the trial. Vital status was determined at subsequent censuses. The primary outcome was aggregate all-cause mortality; country-specific rates were assessed in prespecified subgroup analyses.
    "Results: A total of 1533 communities underwent randomization, 190,238 children were identified in the census at baseline, and 323,302 person-years were monitored." P. 1583.
  • "Census periods started in December 2014, August 2015, February 2016, August 2016, and February 2017." P. 1586.

• 12

  "At the end of the trial, the annual mortality rate for eligible children in the three countries combined was 14.6 deaths per 1000 person-years in communities that received azithromycin (9.1 per 1000 person-years in Malawi, 22.5 in Niger, and 5.4 in Tanzania) and 16.5 deaths per 1000 person-years in communities that received placebo (9.6 per 1000 person-years in Malawi, 27.5 in Niger, and 5.5 in Tanzania). Community-level, intention-to-treat analysis showed that over all four intercensal periods, mortality was 13.5% lower overall (95% confidence interval [CI], 6.7 to 19.8) in the azithromycin group than in the placebo group (P<0.001)." Keenan et al. 2018, p. 1588.

• 13

  "The prespecified primary outcome was the community-level, aggregate, three-country mortality rate determined with the use of data from twice-yearly censuses." Keenan et al. 2018, p. 1585.

• 14

  "Only key trial personnel were aware of which letters corresponded to each group. Participants, observers, investigators, and data-cleaning team members were unaware of the group assignments. Centralized randomization and simultaneous assignment of communities facilitated complete concealment of the assignments. The placebo contained the vehicle of the oral azithromycin suspension and was bottled and labeled identically to azithromycin." Keenan et al. 2018, p. 1584.

• 15

  "MORDOR ClinicalTrials.gov number, NCT02047981." Keenan et al. 2018, p. 1583.

• 16

  "The proportion of children whose census status was recorded as moved or unknown did not differ significantly between the groups (P=0.71 and P=0.36, respectively)." Keenan et al. 2018, p. 1588. These differences are small in magnitude as well (11.63% in azithromycin vs. 11.35% in placebo). Calculations based on Table S3 in Keenan et al. 2018, Supplementary appendix, p. 4.

• 17

  "A total of 66 404 individuals were identified at baseline in the 48 subkebeles participating in the study, including 18 415 children aged 1 to 9 years (May 2006). … The primary prespecified mortality outcome was the comparison of the mortality risk among children aged 1 to 9 years in the treatment vs the control groups using clustered logistic regression; this procedure yielded an odds ratio for mortality in the treatment group of 0.51 (95% confidence interval [CI], 0.29-0.90; P = .02) compared with the control group." Porco et al. 2009, pp. 965-966.

• 18

  "We conducted post hoc analyses using other age ranges. Negative binomial regression of mortality rates in children aged f1 to 5 years yielded a relative mortality rate a factor of 0.47 lower in the treatment groups than in the untreated group (95% CI, 0.26-0.84; P = .01)." Porco et al. 2009, p. 966.

• 19

  O'Brien et al. 2018:

  • "This study was part of the Partnership for the Rapid Elimination of Trachoma (PRET) cluster-randomized trials in The Gambia, Niger, and Tanzania. The Niger trial enrolled participants from 48 grappes (communities) in 6 Centres de Santé Intégrées (CSI) within the Matameye District in the Zinder Region. Detailed eligibility criteria for participation in the Niger trial have been previously reported." P. 1082.
  • "In May 2010, 48 communities in Niger were randomized to receive either annual or biannual mass azithromycin and were monitored until September 2013." P. 1084.
  • "This study employed a 2×2 factorial design to examine both treatment frequency and treatment coverage. Communities were randomized into 4 arms of 12 communities each: 1) annual azithromycin treatment at standard (80%) coverage, (2) annual treatment at enhanced (90%) coverage, 3) biannual treatment at standard (80%) coverage and 4) biannual treatment at enhanced (90%) coverage. Standard coverage (80%) was chosen as a target for the original trial in line with the World Health Organization's recommendation for coverage in trachoma control programs. In addition, the randomization was stratified and blocked based on trachoma prevalence in children, as has been described previously. This report compares the 2 annual treatment arms to the 2 biannual treatment arms." P. 1083.

• 20

  "Among children 6 months to less than 5 years of age, 231 deaths occurred in the annual arm and 173 deaths occurred in the biannual arm during the 3-year study period. The mortality rate was 35.3 deaths per 1000 person-years (95% CI: 30.9 to 40.2) among children in the annual treatment arm and 28.9 deaths per 1000 person-years (95% CI: 24.8 to 33.6) among children in the biannual treatment arm (Table 3). The mortality rate ratio comparing children 6 months to less than 5 years old in the biannual arm to the annual arm was 0.81 (95% CI: 0.66 to 1.00; P = 0.07, Table 3)." O'Brien et al. 2018, p. 1083.

• 21

  "The household census was repeated in May 2015 and in May 2016 to recruit additional eligible children and to detect any deaths that had been missed through the surveillance system. Each year, children who were still younger than 60 months of age on August 1 remained in follow-up for the subsequent trial year, and children who had reached 5 years of age on or before July 31 exited the trial on that date. Enrollment of children in the trial started on August 25, 2014, in Mali and on August 28, 2014, in Burkina Faso. … Children who were enrolled in the trial received the assigned preventive regimen during the annual peak malaria-transmission season (August to November). The drug combinations were administered in four 3-day cycles, at monthly intervals, for three successive seasons. Infants 3 to 11 months of age received a combined 250 mg of sulfadoxine and 12.5 mg of pyrimethamine plus 75 mg of amodiaquine on day 1 and received 75 mg of amodiaquine on days 2 and 3 (Guilin Pharmaceutical, Shanghai, China). In addition, they were randomly assigned to receive either 100 mg of azithromycin or matching placebo on days 1, 2, and 3 (Cipla, Mumbai, India)." Chandramohan et al. 2019, p. 2198.

• 22

  "In the intention-to-treat analysis, the overall number of deaths and hospital admissions during three malaria-transmission seasons was 250 in the azithromycin group and 238 in the placebo group (events per 1000 child-years at risk, 24.8 vs. 23.5; incidence rate ratio, 1.1; 95% confidence interval [CI], 0.88 to 1.3). ... Among children in Burkina Faso and Mali, the addition of azithromycin to the antimalarial agents used for seasonal malaria chemoprevention did not result in a lower incidence of death." Chandramohan et al. 2019, p. 2197.

• 23
  • However, note that in the case of the PRET Trial, control communities also received azithromycin but at a lower dosage (annual treatment vs. biannual treatment).
  • "We identified three cluster-randomized trials undertaken in 1,608 communities from four countries (e.g., Ethiopia, Malawi, Niger, and Tanzania) that met the inclusion criteria. Azithromycin interventions included annual, biannual, and quarterly MDA to the whole community or to children only (Table 1). Control communities included distribution of matching placebo (MORDOR), delayed azithromycin MDA (TANA), and annual azithromycin MDA (standard of care for trachoma-endemic communities, PRET). Across all studies and all sites, a total of 5,486 deaths were observed over 344,905 person-years (mortality rate 15.9 per 1,000 person-years, 95% CI: 15.3–16.1). In azithromycin-treated communities, the mortality rate was 14.4 per 1,000 person-years (95% CI: 13.9–15.0), compared with 17.0 per 1,000 person-years (95% CI: 16.4–17.6) in untreated communities.
    "Across all studies and countries, in a random effects model taking into account between-study heterogeneity, there was a 14.4% reduction in mortality in communities that received azithromycin MDA versus control communities (pooled IRR: 0.856, 95% CI: 0.783–0.937, P = 0.0007; Figure 1). There was moderate heterogeneity across studies (I² = 22.6%, P = 0.11). The pooled incidence rate difference was 2.9 fewer deaths per 1,000 person-years in azithromycin-treated communities than in placebo communities (95% CI: −5.6 to −0.3 deaths per 1,000 person-years, P = 0.03). The results were robust to exclusion of the PRET-Niger study, in which the control arm included annual mass azithromycin distribution (pooled IRR: 0.865, 95% CI: 0.770–0.972, P = 0.02)." Oldenburg et al. 2019, p. 693.

• 24
  • "This trial did not investigate the mechanism by which azithromycin reduced mortality." Keenan et al. 2018, p. 1590.
  • "It is not clear precisely why azithromycin decreased mortality, although infectious diseases are the leading cause of death in Ethiopian children, in particular pneumonia (28%), diarrhea (20%), and malaria (20%). … Further assessment of the mechanism, generalizability, effects of drug resistance or other adverse outcomes, and cost-effectiveness of antibiotic administration in impoverished rural settings may be needed to provide further insight to guide public health policy." Porco et al. 2009, pp. 967-968.
  • O'Brien et al. 2018 do not explicitly mention not investigating mechanisms, but their main results are similar to Keenan et al. 2018 and Porco et al. 2009.

• 25

  "The average rankings for the mechanisms of mortality reduction, in order of importance, were 1.64 for respiratory infection, 2.00 for diarrhea, 2.97 for malaria, 3.96 for other infectious, and 4.65 for noninfectious. All pairwise comparisons differed significantly (P < 0.05) except for respiratory infection and diarrhea (P = 0.19)." See et al. 2015, p. 1107.

• 26

  "In a random sample of 250 verbal autopsies from each of the three sites, we estimated that 41% of the deaths were due to malaria, 18% to diarrhea or dysentery, and 12% to pneumonia." Keenan et al. 2018, p. 1589.

• 27
  • "Verbal autopsy revealed no cause of death for 30.3% (27/82) deaths of children aged 1 to 9 years. For 40.2% (33/82), the cause of death was attributed to malaria, fever, diarrheal or respiratory causes, and the remainder to other causes. For the 55 individuals whose cause of death was revealed by verbal autopsy, 36% (20/55) were attributed to respiratory causes, 9% (5/55) to diarrheal causes, 15% (8/55) to fever or malaria, and an additional 40% (22/55) were attributed to other causes." Porco et al. 2009, p. 967.
  • "Verbal autopsy results are shown in Table 4. Of the 404 children 6 months to less than 5 years of age who died during the study period, verbal autopsy was available for 362 (89.6%). The most common causes of death in both arms were malaria (45.3%) and diarrhea (24.9%). The distribution of cause of death did not differ between the 2 arms (P = 0.36). The infectious disease mortality rate ratio comparing children who died from infectious causes in the biannual arm to the annual arm was 0.73 (95% CI: 0.57 to 0.94; P = 0.02)." O'Brien et al. 2018, p. 1083.

• 28

  "In the prespecified primary analysis, parasitemia was lower in the azithromycin-treated group at month 12 (mean prevalence 8.8%, 95% CI 5.1%–14.3%; 51 of 551 children across all communities) and month 24 (mean 3.5%, 95% CI 1.9%–5.5%; 21 of 567 children across all communities) than it was in the placebo-treated group at month 12 (mean 15.3%, 95% CI 10.8%–20.6%; 81 of 548 children across all communities) and month 24 (mean 4.8%, 95% CI 3.3%–6.4%; 28 of 592 children across all communities) (P = 0.02). Communities treated with azithromycin had approximately half the odds of parasitemia compared to those treated with placebo (odds ratio [OR] 0.54, 95% CI 0.30 to 0.97). Parasite density was lower in the azithromycin group than the placebo group at 12 and 24 months (square root–transformed outcome; density estimates were 7,540 parasites/μl lower [95% CI −350 to −12,550 parasites/μl; P = 0.02] at a mean parasite density of 17,000, as was observed in the placebo arm)." Arzika et al. 2019, p. 2.

• 29
  • "Mass distribution of a single dose of oral azithromycin for trachoma is associated with a significant short-term reduction in ALRI [acute lower respiratory infection] morbidity among young children." Coles et al. 2012, p. 341.
  • "Mass administration of azithromycin to eliminate blindness due to trachoma has raised concerns regarding the emergence of antimicrobial resistance. During 2000, we compared the antimicrobial resistance of nasopharyngeal pneumococcal isolates recovered from and the prevalence of impetigo, respiratory symptoms, and diarrhea among 458 children in Nepal before and after mass administration of azithromycin. No azithromycin-resistant pneumococci were isolated except from 4.3% of children who had received azithromycin during 2 previous mass treatments (P<.001). There were decreases in the prevalence of impetigo (from 14% to 6% of subjects; adjusted odds ratio [OR], 0.41; 95% confidence interval [CI], 0.21–0.80) and diarrhea (from 32% to 11%; adjusted OR, 0.26; 95% CI, 0.14–0.43) 10 days after azithromycin treatment. The absence of macrolide-resistant isolates after 1 mass treatment with azithromycin is encouraging, although the recovery of azithromycin-resistant isolates after 2 mass treatments suggests the need for resistance monitoring when multiple rounds of antimicrobial treatment are given." Fry et al. 2002, p. 395.

• 30

  "There are several possible explanations for the different outcomes of these two trials. One possible explanation is that azithromycin, which has antimalarial activity, contributed to decreased mortality in the MORDOR trial partly through its effect on malaria, and this benefit was lost when an additional, effective antimalarial combination was given at the same time as azithromycin. However, the effect of azithromycin on malaria has been inconsistent when azithromycin has been given in mass drug administration programs. In addition, all the children in our trial received sulfadoxine–pyrimethamine, which has weak antimicrobial properties, and this may have reduced the potential benefit of adding another antimicrobial to the regimen. Finally, coverage with a pneumococcal conjugate vaccine was high among the children in our trial, and this may have reduced the potential benefit of azithromycin in lowering mortality from pneumonia." Chandramohan et al. 2019, p. 2205.

• 31

  "Mortality rates were 5.7% lower (95% CI, −9.7 to 18.9) in the azithromycin group than in the placebo group in Malawi (P=0.45), 18.1% lower (95% CI, 10.0 to 25.5) in Niger (P<0.001), and 3.4% lower (95% CI, −21.2 to 23.0) in Tanzania (P=0.77). … Efficacy did not differ significantly between the two groups by country (P=0.17), age group (P=0.20), treatment period (P=0.09), or treatment coverage (P=0.34)." Keenan et al. 2018, pp. 1588-1589.

• 32

  "We examined whether baseline mortality risk, as a function of child age and site, modified the azithromycin mortality-reduction effect in the Macrolide Oraux pour Réduire les Décès avec un Oeil sur la Résistance (MORDOR) clinical trial. We used the Cox proportional hazards model with an interaction term. Three models were examined representing three sources for the baseline-risk covariate: two using sources external to MORDOR and the third leveraging data within MORDOR. All three models provided moderate evidence for the effect becoming stronger with increasing baseline mortality (P = 0.02, 0.02, and 0.07, respectively) at the rate of approximately 6–12% additional mortality reduction per doubling of baseline mortality. Etiological and programmatic implications of these findings are discussed." Oron et al. 2019, p. 1.

• 33

  "Here, we evaluated the relationship between the underlying mortality rate and the efficacy of azithromycin for the prevention of child mortality using data from multiple sites in Ethiopia, Malawi, Niger, and Tanzania. Between regions, we find no strong evidence of effect modification of the efficacy of azithromycin distribution for the prevention of child mortality by the underlying mortality rate (P = 0.12), although a modest effect is consistent with our findings. Higher mortality settings could be prioritized, however, because of the larger number of deaths which could be averted with azithromycin distribution." Porco et al. 2019, p.1.

• 34

  "A perfectly proportional azithromycin effect (i.e., no effect modification) would trace a diagonal linear pattern from the origin. By contrast, the observed pattern does not substantially depart from zero effect until a mortality rate of ∼10 deaths/1,000 py. At the high-mortality end, the effect increases nonlinearly. The latter pattern is driven by Niger data only, as this was the only site with observed age-bin placebo mortality rates greater than 25 deaths/1,000 py. However, observations from all three sites overlap in the range of 10–25 deaths/1,000 py, where they all seem to follow a similar trend." Oron et al. 2019, p. 3.

• 35

  "Children in the youngest age group (1 to 5 months of age) had the highest overall mortality and the largest observed difference in mortality with azithromycin as compared with placebo (24.9% lower with azithromycin; 95% CI, 10.6 to 37.0; P=0.001)." Keenan et al. 2018, p. 1588.

• 36

  "Efficacy did not differ significantly between the two groups by country (P=0.17), age group (P=0.20), treatment period (P=0.09), or treatment coverage (P=0.34)." Keenan et al. 2018, p. 1589.

• 37
  • See Figure 4, Keenan et al. 2018, p. 1589.
  • "Efficacy did not differ significantly between the two groups by country (P=0.17), age group (P=0.20), treatment period (P=0.09), or treatment coverage (P=0.34)." Keenan et al. 2018, p. 1589.

• 38

  "Among children 6 months to less than 5 years of age, 231 deaths occurred in the annual arm and 173 deaths occurred in the biannual arm during the 3-year study period. The mortality rate was 35.3 deaths per 1000 person-years (95% CI: 30.9–40.2) among children in the annual treatment arm and 28.9 deaths per 1000 person-years (95% CI: 24.8–33.6) among children in the biannual treatment arm (Table 3). The mortality rate ratio comparing children 6 months to less than 5 years old in the biannual arm to the annual arm was 0.81 (95% CI: 0.66–1.00; P = 0.07, Table 3)." O'Brien et al. 2018, p. 1082.

• 39

  "Macrolides: e.g. azithromycin, clarithromycin, erythromycin: These antibiotics are considered highest-priority critically important antimicrobials on the CIA List and carry a high risk of selection of bacterial resistance (particularly resistance to macrolides). With its remarkably long half-life, azithromycin carries the highest risk of resistance among the macrolides. Azithromycin is listed on the EML/EMLc as a first-choice option for trachoma, yaws, Chlamydia trachomatis, cholera and Neisseria gonorrhoeae, and as a second-choice option for acute invasive bacterial diarrhoea/dysentery and Neisseria gonorrhoeae. Clarithromycin is listed as a first-choice option for Helicobacter pylori and community-acquired pneumonia (severe), and as a second-choice option for pharyngitis." WHO, The Selection and Use of Essential Medicines, 2017, p. 65.

• 40

  "The most commonly studied organism was S pneumoniae, with 12 studies (including the TEF trial) reporting resistance in isolates of this bacterium after mass azithromycin distribution for trachoma treatment. S pneumoniae is an important commensal organism that colonises the nasopharynx and can cause pneumonia. Overall, these studies showed an increase in macrolide resistance in S pneumoniae immediately after treatment, which appears to dissipate with time. Five studies (including the TEF trial) included an untreated control group and showed a substantially increased resistance in S pneumoniae in communities that received azithromycin compared with those that did not. This finding indicates that the increase is probably due to the azithromycin intervention rather than secular trends. Although heterogeneity in study design, setting, population, treatment frequency, and follow-up time precluded formal meta-analysis, trends in the included studies suggested that increasing treatment frequency (eg, single, annual, and biannual) increased selection for macrolide resistance in S pneumoniae. Studies and programmes that are planning greater frequencies of azithromycin distribution should consider the potential for increased selection for macrolide resistance.
  "Some evidence of selection for macrolide resistance following mass azithromycin distribution was noted in other organisms, including E coli and S aureus. Enterotoxigenic E coli strains are a major cause of childhood diarrhoea, although macrolides are not typically used against them." O'Brien et al. 2019, p. 22.

• 41
  • "The proportion of macrolide resistance in nasopharyngeal S. pneumoniae at the community level was higher in the communities receiving azithromycin (mean, 12.3%; 95% CI, 5.7 to 20.0) than in the those receiving placebo (mean, 2.9%; 95% CI, 0 to 6.1; P=0.02). (Details are provided in Table 1, and in Table S1 in the Supplementary Appendix.) Similarly, determinants of macrolide resistance in the intestinal flora were more prevalent in the communities that received azithromycin (68.0%; 95% CI, 61.3 to 74.0) than in those that did not (46.7%; 95% CI, 36.0 to 54.0; P=0.002)." Doan et al. 2019a, p. 2272.
  • "The MORDOR I trial, conducted in Niger, Malawi and Tanzania, demonstrated that mass azithromycin distribution to preschool children reduced childhood mortality. However, the large but simple trial design precluded determination of the mechanisms involved. Here we examined the gut microbiome of preschool children from 30 Nigerien communities randomized to either biannual azithromycin or placebo. Gut microbiome γ-diversity was not significantly altered (P = 0.08), but the relative abundances of two Campylobacter species, along with another 33 gut bacteria, were significantly reduced in children treated with azithromycin at the 24-month follow-up. Metagenomic analysis revealed functional differences in gut bacteria between treatment groups. Resistome analysis showed an increase in macrolide resistance gene expression in gut microbiota in communities treated with azithromycin (P = 0.004). These results suggest that prolonged mass azithromycin distribution to reduce childhood mortality reduces certain gut bacteria, including known pathogens, while selecting for antibiotic resistance." Doan et al. 2019b, p. 1370.

• 42

  "Mortality rates in the two treated groups are shown according to intercensal period (Figure 2) and according to year (Table 2). A total of 24.0 deaths (95% confidence interval [CI], 22.1 to 26.3) per 1000 person-years were recorded in communities that were receiving the first year of azithromycin distribution, and 23.3 deaths (95% CI, 21.4 to 25.5) per 1000 person-years were recorded in communities receiving the third year of azithromycin distribution. We found no evidence that the first year of treatment had a greater effect on mortality than the third year of treatment, with 3.5% (95% CI, −8.3 to 14) more deaths in communities receiving the first year of treatment (P=0.55) (Table 2)." Keenan et al. 2019, p. 2210.

• 43

  See our cost-effectiveness analysis of mass distribution of azithromycin, “Azithromycin” sheet, “Azithromycin vs cash” row, “Preferred” column.